An official website of the United States government
Background: At the time of cancer diagnosis, it is crucial to accurately classify malignant gastric tumors and the possibility that patients will survive. Objective: This study aims to investigate the feasibility of identifying and applying a new feature extraction technique to predict the survival of gastric cancer patients. Methods: A retrospective dataset including the computed tomography (CT) images of 135 patients was assembled. Among them, 68 patients survived longer than three years. Several sets of radiomics features were extracted and were incorporated into a machine learning model, and their classification performance was characterized. To improve the classification performance, we further extracted another 27 texture and roughness parameters with 2484 superficial and spatial features to propose a new feature pool. This new feature set was added into the machine learning model and its performance was analyzed. To determine the best model for our experiment, Random Forest (RF) classifier, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Naïve Bayes (NB) (four of the most popular machine learning models) were utilized. The models were trained and tested using the five-fold cross-validation method. Results: Using the area under ROC curve (AUC) as an evaluation index, the model that was generated using the new feature pool yields AUC = 0.98 ± 0.01, which was significantly higher than the models created using the traditional radiomics feature set (p < 0.04). RF classifier performed better than the other machine learning models. Conclusions: This study demonstrated that although radiomics features produced good classification performance, creating new feature sets significantly improved the model performance.
more »
« less
Improving plankton image classification using context metadata
Abstract Advances in both hardware and software are enabling rapid proliferation of in situ plankton imaging methods, requiring more effective machine learning approaches to image classification. Deep Learning methods, such as convolutional neural networks (CNNs), show marked improvement over traditional feature‐based supervised machine learning algorithms, but require careful optimization of hyperparameters and adequate training sets. Here, we document some best practices in applying CNNs to zooplankton and marine snow images and note where our results differ from contemporary Deep Learning findings in other domains. We boost the performance of CNN classifiers by incorporating metadata of different types and illustrate how to assimilate metadata beyond simple concatenation. We utilize both geotemporal (e.g., sample depth, location, time of day) and hydrographic (e.g., temperature, salinity, chlorophylla) metadata and show that either type by itself, or both combined, can substantially reduce error rates. Incorporation of context metadata also boosts performance of the feature‐based classifiers we evaluated: Random Forest, Extremely Randomized Trees, Gradient Boosted Classifier, Support Vector Machines, and Multilayer Perceptron. For our assessments, we use an original data set of 350,000 in situ images (roughly 50% marine snow and 50% non‐snow sorted into 26 categories) from a novel in situZooglider. We document asymptotically increasing performance with more computationally intensive techniques, such as substantially deeper networks and artificially augmented data sets. Our best model achieves 92.3% accuracy with our 27‐class data set. We provide guidance for further refinements that are likely to provide additional gains in classifier accuracy.
more »
« less
- Award ID(s):
- 1637632
- PAR ID:
- 10372162
- Publisher / Repository:
- Wiley Blackwell (John Wiley & Sons)
- Date Published:
- Journal Name:
- Limnology and Oceanography: Methods
- Volume:
- 17
- Issue:
- 8
- ISSN:
- 1541-5856
- Page Range / eLocation ID:
- p. 439-461
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Analysis of ion-kinetic instabilities in solar wind plasmas is crucial for understanding energetics and dynamics throughout the heliosphere, as evident from spacecraft observations of complex ion velocity distribution functions (VDFs) and ubiquitous ion-scale kinetic waves. In this work, we explore machine learning (ML) and deep learning (DL) classification models to identify unstable cases of ion VDFs driving kinetic waves. Using 34 hybrid particle-in-cell simulations of kinetic protons andα-particles initialized using plasma parameters derived from solar wind (SW) observations, we prepare a data set of nearly 1600 VDFs representing stable/unstable cases and associated plasma and wave properties. We compare feature-based classifiers applied to VDF moments, such as support vector machine and random forest (RF), with DL convolutional neural networks (CNNs) applied directly to VDFs as images in the gyrotropic velocity plane. The best-performing classifier, RF, has an accuracy of 0.96 ± 0.01, and a true skill score of 0.89 ± 0.03, with the majority of missed predictions made near stability thresholds. We study how the variations of the temporal derivative thresholds of anisotropies and magnetic energies, and sampling strategies for simulation runs, affect classification. CNN-based models have the highest accuracy of 0.88 ± 0.18 among all considered if evaluated on the runs entirely not used during the model training. The addition of theE⊥power spectrum as an input for the ML models leads to the improvement of instability analysis for some cases. The results demonstrate the potential of ML and DL for the detection of ion-scale kinetic instabilities using spacecraft observations of SW and magnetospheric plasmas.more » « less
-
Abstract Despite forming under different flow conditions, the geometries of tidal and fluvial channel planforms and planform transformations display significant overlap, hindering efforts to differentiate them geometrically. Although studies have demonstrated that globally, tidal and fluvial planforms are statistically distinct based on meander metrics, there are currently no machine‐learning methodologies for classifying channels as tidal or fluvial that do not focus on meander‐specific geometries. In this study, we present a methodology for classifying channel planforms as tidal or fluvial using statistical representations of channel planforms and machine‐learning algorithms. Using a dataset of 4294 tidal and fluvial channel segments (63 channel reaches), we trained three machine‐learning classifiers (Logistic Regression, Multi‐layer Perceptron, and Random Forest) across 69 trials to identify the machine‐learning algorithm and variables that perform best at classifying channel reaches. We evaluated the performance of the classifiers at three thresholds based on the percent of channel segments correctly identified in a given reach (>50%, >66% and >75%). At the >50% classification threshold, all three classifiers attained a 95% reach‐scale accuracy during individual trials. However, at higher classification thresholds, the RF classifier performed best. Feature importances from the RF classifier indicate that measures of the central tendency and minimum/maximum of the normalized radius of curvature convolved with normalized width of channel segments play a key role in differentiating between the planforms, with normalized width also contributing to the difference. This indicates that the relationship between width and radius of curvature is more important than width or measures of curvature on their own. This result likely reflects the downstream funnelling of tidal channels and the limitation on the sharpness of bends associated with increased width. These methods have potential for application in the study of channels preserved on relict geomorphic surfaces and in mixed‐energy settings.more » « less
-
Hadsell, R; Richards, B; Zador, A (Ed.)Deep neural network modeling of biological visual processing is widespread: brains are archetypal pattern analyzers and deep CNNs are currently the best object classifiers. Implicit is the assumption that cortex can be well approximated by CNNs, from which it follows that CNNs are an appropriate foundation for AI. We examine whether this approximation holds using a novel neural manifold obtained with machine learning techniques.more » « less
-
Zelinski, Michael E.; Taha, Tarek M.; Howe, Jonathan (Ed.)Image classification forms an important class of problems in machine learning and is widely used in many realworld applications, such as medicine, ecology, astronomy, and defense. Convolutional neural networks (CNNs) are machine learning techniques designed for inputs with grid structures, e.g., images, whose features are spatially correlated. As such, CNNs have been demonstrated to be highly effective approaches for many image classification problems and have consistently outperformed other approaches in many image classification and object detection competitions. A particular challenge involved in using machine learning for classifying images is measurement data loss in the form of missing pixels, which occurs in settings where scene occlusions are present or where the photodetectors in the imaging system are partially damaged. In such cases, the performance of CNN models tends to deteriorate or becomes unreliable even when the perturbations to the input image are small. In this work, we investigate techniques for improving the performance of CNN models for image classification with missing data. In particular, we explore training on a variety of data alterations that mimic data loss for producing more robust classifiers. By optimizing the categorical cross-entropy loss function, we demonstrate through numerical experiments on the MNIST dataset that training with these synthetic alterations can enhance the classification accuracy of our CNN models.more » « less
